Categorization is among the most important cognitive skills that humans possess. It allows us to navigate in a dangerous world, and to find food, shelter, and friends. The evidence is now overwhelming that humans have multiple category-learning systems, which are largely neuroanatomically separate, learn by qualitatively different rules, and have adapted to learning different types of category structures. A natural next question to investigate is how these various systems interact. This is an important problem because during daily life we must often switch between different categorization systems (e.g., explicit and procedural). For example, many components of driving are procedural, but at the same time some explicit decisions are required. Following an explicit decision, a failure to quickly switch back to a procedural strategy could greatly increase the risk of an accident. The proposed research studies how explicit and procedural category learning are coordinated and how control is transferred between these systems. We take an integrative, cross-disciplinary, converging operations approach that combines behavioral, neuropsychological (with Parkinson's disease patients), functional magnetic resonance imaging, and transcranial magnetic stimulation studies with the goal of building and testing biologically detailed computational models of the brain circuits that mediate categorization and system-switching behavior. This proposal is to continue a program that (a) provided much of the existing evidence that humans have multiple category-learning systems, (b) mapped out the neural networks that mediate each system, and (c) discovered many unique properties of these systems. During the previous period (2R01 MH3760), we made significant progress in several areas. One was to understand how learning in the various systems is coordinated. Toward this end, we reported evidence that trial-by-trial switching between explicit and procedural categorization strategies is extremely difficult. The proposed research, which continues our investigations of system interactions, has three aims. Aim 1 is to identify the cognitive components of system switching. Aim 2 is to identify the neural basis of system switching, and Aim 3 is to develop and test a biologically detailed computational model of system switching. The model we develop should be able to provide accurate accounts of all data from Aims 1 and 2, as well as data from various published single-unit recording studies. In addition, the model will make specific predictions about how drugs, genes, and focal lesions should affect behavior and it will make novel predictions about behavioral and pharmacological interventions that might improve system switching in category learning.